Plot passes null to arrayLike

[mwaibel-go]

I’m not sure if this is a bug or if I missed something, so I’m writing this as a question: Why does the following chart config throw an error?

Plot.plot({
  x: { type: 'time', },
  marks: [
    Plot.tickX(
      [new Date('2026-01-01'), new Date('2026-01-03')],
      { transform: (data) => data }
    )
  ]
})

TypeError: right-hand side of 'in' should be an object, got null

[mohammadrezwankhan]

The Date values and x: {type: "time"} are fine. The problem is the mark-level transform contract.

In Plot’s Mark.initialize, a mark transform is destructured as {data, facets}. Your (data) => data returns the bare array, so Plot ends up with data === undefined; the later arrayLike/in error is just the downstream symptom. The docs state the same contract: a custom transform receives (data, facets, options) and must return a {data, facets} object (https://observablehq.com/plot/features/transforms#transformoptions-transform).

You do not need a transform for this example. tickX defaults its x channel to the identity accessor, so this is enough:

const data = [
  new Date("2026-01-01"),
  new Date("2026-01-03")
];

Plot.plot({
  x: {type: "time"},
  marks: [Plot.tickX(data)]
});

If you intentionally want a mark transform that leaves the data unchanged, preserve both values:

Plot.tickX(data, {
  transform: (data, facets) => ({data, facets})
});

If what you meant was an x-channel accessor, put the function on x instead of using the mark transform:

Plot.tickX(data, {x: d => d});

For filtering or reshaping rows, use a built-in transform such as Plot.filter or return the correctly shaped object from a custom transform. The scale type is not the cause here; returning an array from the mark-level transform is.

[mwaibel-go]

Thank you for your thorough answer. I was indeed confusing channel accessors and mark transforms.

The example given in Mark options looks very similar and I didn’t read it right.

Plot.dot(numbers, {x: {transform: (data) => data}})